摘要
滑坡与其评价因子之间的关联计算将直接影响滑坡易发性评价结果的精度,因此需要研究不同关联模型对滑坡易发性评价的影响。以白鹤滩库区象鼻岭-江边村段库岸段为研究区,选取以层次分析法(AHP)为代表的经验模型和以信息量法(IV)为代表的统计模型分别与BP神经网络(BPNN)模型耦合进行滑坡易发性评价,同时将仅采用BPNN模型的评价结果作对比。采用ROC曲线、频率比、易发性指数分布(均值、标准差)等方式对滑坡易发性结果进行对比评价。结果显示,IV-BPNN耦合模型滑坡易发性评价结果精度高于AHP-BPNN耦合模型,统计分析模型在滑坡评价因子关联分析上表现更为优秀;单一BPNN模型易发性评价结果整体精度低于耦合模型。在进行滑坡易发性评价时采用评价因子关联法的耦合模型优势较为明显,并且以信息量法为代表的统计分析模型作评价因子关联法进行耦合计算,获得的易发性评价结果精度最高、可靠性最强,评价结果与研究区实际的滑坡分布更为相符。
The correlation calculation model between landslide and assessment factors has a direct impact on the accuracy of landslide susceptibility evaluation.From Baihetan reservoir area the impact of various correlation models on landslide susceptibility evaluation is studied.The empirical model represented by the analytic hierarchy process(AHP)and the statistical model represented by the information value(IV)are respectively coupled with BP neural network(BPNN)model to evaluate landslide susceptibility.The evaluation results of the BPNN model alone are used in the comparison.To compare and assess landslide susceptibility data,the ROC curve,frequency ratio,and susceptibility index distribution(mean value and standard deviation)are utilized.The IVBPNN coupling model beats the AHP-BPNN coupling model in terms of accuracy,and the statistical analysis model beats the AHP-BPNN coupling model in terms of landslide evaluation factor correlation analysis.Susceptibility evaluation results from a single BPNN model are less accurate than those from a coupled model.The study found that using the coupling model of the evaluation factor correlation approach to assess landslide susceptibility has evident benefits.The statistical analysis model represented by the information method is used as an evaluation factor for the coupling calculation,and the sensitivity evaluation result obtained by the correlation method is more accurate and reliable.It is more in line with the actual distribution of landslides in the study area.
作者
薛正海
冯文凯
韩靖楠
易小宇
吴明堂
XUE Zheng-hai;FENG Wen-kai;HAN Jing-nan;YI Xiao-yu;WU Ming-tang(State Key Laboratory of Geohazard Prevention and Geoenvironment Protection,Chengdu University of Technology,Chengdu 610059,China;Zhejiang Huadong Construction Engineering Co.,Ltd.,Hangzhou 310014,China)
出处
《桂林理工大学学报》
CAS
北大核心
2022年第4期809-819,共11页
Journal of Guilin University of Technology
基金
国家自然科学基金项目(41977252)
地质灾害防治与地质环境保护国家重点实验室自主探索课题(SKLGP2020Z001)
浙江华东建设工程有限公司科研项目(KY2020-HDJS-19)。
关键词
滑坡易发性
关联分析
层次分析模型
信息量模型
BP神经网络模型
landslide susceptibility
correlation analysis
analytic hierarchy process model
information value model
back propagation neural network model
